Updating/overwriting attribute with rlm_python

I’m using rlm_python, returning a tuple to update attributes in the reply. I’ve noticed that if the request already has a value for that attribute, it is not overwritten. That is, my operation becomes a no-op. I suppose this matches the behavior of the “=“ operator. Is there a way to behave like the “:=“ operator, overwriting the existing value? Or should I strip the value prior to invoking my python module? Thanks, Norman